What about an index on an I-DESC? (Does universe do that?) -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Mark Johnson Sent: Friday, October 21, 2005 5:03 PM To: u2-users@listserver.u2ug.org Subject: Re: [U2] Double Indexes
I putzed around with correlative-based indexed fields and got a lot of errmsg static. I'll try your suggestions. I'm trying to use the system instead of maintain manual xref files. This system is crammed full of 2M, 3M, 6Million record files and there is a lot of NUM1:NUM2 and NUM2:NUM1 (example) cross referencing done manually but none for DATE/CUST. I thought of creating a single Date/Cust xref file that could grab the eventual 800 average keys and let the secondary process filter it down to the desired subset but again, weaving it into this scrappy application (80% TPH) would guarantee not being perfect. Thanks. ----- Original Message ----- From: "Scott Ballinger" <[EMAIL PROTECTED]> To: <u2-users@listserver.u2ug.org> Sent: Friday, October 21, 2005 11:42 AM Subject: RE: [U2] Double Indexes > [snip] > >Does U2 support indexing virtual fields, ie DATE.CUST? I don't think D3 > can. > > Sure, both U2 and D3 support virtual index fields. > > U2 I type: > CUST:".":DATE > > One caveat I know of is that in D3 you must use A-correlatives to define > the index, e.g. > > Cust.date: > a5:".":6 > Or > A0(call cust.date) > > Create-index myfile cust.date > > Select myfile with cust.date eq "12345.]" -> all records with cust = > 12345 > Select myfile with cust.date eq "12345.13809" -> cust 12345 and date = > 10-21-05 > > However, the fact that the date in cust.date must be specified in > internal format reduces the utility of this approach, as does the fact > that indexed dates can't be selected using greater than or less than > type logic. > > Selecting by customer and/or date is pretty common; I usually end up > creating a separate xref index file with the key = cust dot date and a > mv list of myfile ids in attribute 1. Then you can > > select xref.file with cust = "12345" and with date ge "10-1-05" and le > "10-20-05" myfile.ids > > This is usually pretty fast in itself, and you could always index cust > and date in xref.file to make it faster under some circumstances. > > /Scott Ballinger > Pareto Corporation > Edmonds WA USA > 206 713 6006 > > P.s. I did not know that that UV10 supports multiple index queries - > very cool, and good to know! > ------- > u2-users mailing list > u2-users@listserver.u2ug.org > To unsubscribe please visit http://listserver.u2ug.org/ ------- u2-users mailing list u2-users@listserver.u2ug.org To unsubscribe please visit http://listserver.u2ug.org/ ------- u2-users mailing list u2-users@listserver.u2ug.org To unsubscribe please visit http://listserver.u2ug.org/
